Characterization of ligand binding of a soluble human insulin-like growth factor I receptor variant suggests a ligand-induced conformational change.

Details of the signal transduction mechanisms of the tyrosine kinase family of growth factor receptors remain elusive. In this work, we describe an extensive study of kinetic and thermodynamic aspects of growth factor binding to a soluble extracellular human insulin-like growth factor-I receptor (sIGF-IR) variant. The extracellular receptor domains were produced fused to an IgG-binding protein domain (Z) in transfected human 293 cells as a correctly processed secreted α-β′-Z dimer. The receptor was purified using IgG affinity chromatography, rendering a pure and homogenous protein in yields from 1 to 5 mg/liter of conditioned cell media. Biosensor technology (BIAcore) was applied to measure the insulin-like growth factor-I (IGF-I), des(1-3)IGF-I, insulin-like growth factor-II, and insulin ligand binding rate constants to the immobilized IGF-IR-Z. The association equilibrium constant, Ka, for the IGF-I interaction is determined to 2.8 × 108 M−1 (25 °C). Microcalorimetric titrations on IGF-I/IGF-IR-Z were performed at three different temperatures (15, 25, and 37 °C) and in two different buffer systems at 25 °C. From these measurements, equilibrium constants for the 1:1 (IGF-I:(α-β′-Z)2) receptor complex in solution are deduced to 0.96 × 108 M−1 (25 °C). The determined heat capacity change for the process is large and negative, −(1)0.51 kcal (K mol)−1. Further, the entropy change (ΔS) at 25 °C is large and negative. Far- and near-UV circular dichroism measurements display significant changes over the entire wavelength range upon binding of IGF-I to IGF-IR-Z. These data are all consistent with a significant change in structure of the system upon IGF-I binding.

Cellular receptor structure and function relationships have become the focus of an increasing amount of research, relevant for biological understanding as well as for possible pharmaceutical applications. The insulin-like growth factor-I receptor (IGF-I R ) 1 is a transmembrane glycoprotein and belongs to the receptor tyrosine kinase family that includes, among others, the insulin (ins R ), epidermal growth factor, and platelet-derived growth factor receptors (1). Insulin-like growth factor-I (IGF-I) is a major regulator of both cellular growth and metab-olism (2). The components of the IGF hormone signaling system are tightly regulated, both in tissue-specific and developmentally specific manners. The IGF molecular system contains three known peptide ligands (IGF-I, IGF-II, and insulin), three types of cellular receptors (IGF-I R , ins R , and IGF-II/mannose 6-phosphate), and six known subtypes of circulating IGF-binding proteins (IGFBPs) (IGFBP-1 to IGFBP-6). The molecular components of the IGF signaling system have been the subject of several recent review articles (3,4). The mechanism by which insulin and IGF molecules interact with their respective receptor to mediate signaling is still not understood in great detail at the molecular level. The ligand-induced receptor activation has been suggested to involve a conformational switch in the quaternary structure upon ligand binding, with movements of the extracellular ␣ parts and a congregation of the cytoplasmic tyrosine kinase regions to enable activation (5,6). However, the complexity of IGF-I signaling remains rather elusive almost 10 years after the molecular cloning of the human IGF-I receptor gene (7).
IGF-I R is active as a preformed heterotetrameric receptor containing two extracellular ␣-domains with ligand binding sites and two transmembrane-spanning ␤-domains, also harboring the cytoplasmic tyrosine kinase activity (Fig. 1A). The chains are disulfide-linked in a ␤-␣-␣-␤ arrangement with possibly three disulfide linkages between the ␣ chains and one disulfide bond between each ␣-␤ chain pair (8). The gene is transcribed and translated as a single ␣-␤ polypeptide chain and post-translationally processed and assembled into a mature receptor. It has been shown for the homologous insulin receptor that receptor maturation is achieved through proteolytic processing, oligosaccharide attachment, and disulfidelinked homodimerization during the transport from endoplasmic reticulum to Golgi and further to the cell surface (9).
In this paper, the ligand binding properties of a purified extracellular portion of the IGF-I R have been characterized using BIAcore TM biosensor analysis. From the BIAcore data, molecular association and disassociation rates as well as binding equilibrium constants were calculated. Binding properties of the soluble IGF-I R receptor variant were analyzed for the ligands IGF-I, des(1-3)-IGF-I, IGF-II, and insulin, respectively. Near-and far-UV circular dichroism measurements were performed to establish an estimation of possible secondary structure changes upon ligand association.
Microcalorimetric titration binding experiments of IGF-I to the heterotetrameric IGF-I R -Z were performed at three different temperatures, from which enthalpies, ⌬H 0 , Gibbs free energies, ⌬G 0 , entropies, ⌬S 0 , and heat capacity, ⌬C p 0 , of the binding process could be calculated. In this work, we have used the model proposed by Freire et al. (10 -13) to analyze microcalorimetric titration data of the binding of IGF-I to the het-erotetrameric IGF-I R -Z. From this analysis, the change in hydration-accessible polar and nonpolar surface areas were calculated, as well as the entropic contribution to changes in conformational degree of freedom upon binding. This is to our knowledge the first combined study, using both BIAcore and titration calorimetry analyses of ligand binding, of a growth factor receptor from the insulin-like growth factor family of tyrosine kinase receptors.

EXPERIMENTAL PROCEDURES
Ligand Proteins-Native IGF-I, des(1-3)IGF-I, and IGF-II were produced in Escherichia coli and purified as described previously (14,15). Recombinant human insulin was purchased from Sigma.
Production Vector Construction-A vector was constructed for the production of IGF-I R including an affinity handle for purification and analysis purposes. Human IGF-I receptor cDNA was cloned using linked reverse transcriptase-polymerase chain reaction of placental total RNA (Clontech) into a vector containing cytomegalovirus promoter and enhancer regions, a single Z domain gene, a stop codon followed by SV40 poly(A)-signal and 3Ј-untranslated regions. A sequence coding for a thrombin-sensitive cleavage linker, SGLVPRGSG, was inserted between the receptor ␤-3Ј and Z-5Ј coding regions at amino acid position 933 according to the numbering of Ullrich et al. (7). The outline of the expression cassette of the resulting plasmid, pKGE978, is shown in Fig. 2.
Transfection and Expression-Human primary kidney 293 cells were transfected for transient expression using the calcium phosphate method (16). Cells were incubated for 24 h in Dulbecco's modified Eagle's medium/McCoy's 5A 1:1 (Life Technologies, Inc.) plus 5% bovine calf serum (HyClone), upon which the medium was replaced by serumfree medium and cells were incubated for another 24 h. Stable transfected cell lines were established from 293 human kidney cells transfected with pKGE978 and co-transfected with plasmid pKGE800 carrying a neomycin resistance gene. Selection was performed using G418, Geneticin, (Sigma) at 1 mg/ml.
Receptor Purification-Cell media harvested after 48 -72 h of growth were centrifuged to remove cell debris and further diluted with 1 ⁄2 volume of water and 1 ⁄10 volume of 10 ϫ TST (1 ϫ TST: 50 mM Tris-HCl, 150 mM NaCl, 0.05% Tween 20, pH 7.4) to buffer the media to a pH suitable for the chromatography step. Diluted and buffered medium was filtered through a 0.22-m filter to remove insoluble material prior to chromatography. Affinity chromatography purification was performed using an IgG-Sepharose FF column (Pharmacia Biotech, Sweden), previously equilibrated with 1 ϫ TST. Unspecifically bound proteins and impurities were removed with 10 column volumes of 1 ϫ TST and 2 volumes of 5 mM ammonium acetate, pH 5.0. Bound fusion protein was subsequently eluted by 0.2 M acetic acid, pH 2.8. Fractions were immediately neutralized through collection in tubes containing 1 ⁄3 fraction volume of 1 M Tris-HCl, pH 8.0. The buffer of the eluted material was exchanged to 10 mM HEPES, pH 7.4, through several rounds of dilution and concentration using a Filtron ultrafiltration unit with a 30-kDa molecular mass cut-off (Filtron). Purified and concentrated receptor was stored in aliquots at Ϫ80°C.
Biochemical Characterization-Quantitative amino acid analysis was performed by acid hydrolysis of the peptide chain in 6 M HCl at 155°C for 45 min, followed by analysis using an ion exchange column and ninhydrin derivative detection. The analysis was performed on a Beckman 6300 amino acid analyzer, equipped with a System Gold data handling system (Beckman).
The N-terminal protein sequence was determined using a HP G1005A protein sequencing system with on-line phenylthiohydantoin analysis, version 2.2 chemistry (Hewlett-Packard).
The oligosaccharide content of the receptor was determined by analysis of hexose content as described by Kenne and Strömberg (17). Briefly, hexose content was determined by acid hydrolysis and alditol acetate derivation of released sugars. Sugar alditol acetates were separated by gas chromatography-mass spectrometry on a fused silica capillary column (25 m ϫ 0.2 mm) at 250°C. An HP 5970B mass selective detector (Hewlett-Packard) was used for monitoring eluted peaks. Samples were compared with calibration curves constructed using defined sugar standards. N-Acetylneuraminic acid (sialic acid) content was analyzed by cleavage of the glycosylating carbohydrate chains with hydrogen chloride in methanol, transferring the sialic acid to its methyl ester methyl glycoside and further separation by gas chromatography. 2 Protein homogeneity was evaluated by SDS-polyacrylamide gel electrophoresis (18) using NOVEX precast 4 -20% gradient polyacrylamide gels (Novel Experimental Technology) stained with Coomassie Blue staining. The Z affinity tail was detected using semidry transfer to nitrocellulose membranes and peroxidase antiperoxidase staining as detailed previously (19). Apparent molecular mass of both free and complexed receptor was determined using laser light-scattering analysis, performed with a Protein Solutions Dyna Pro 801 TC (Protein Solutions). The receptor was analyzed at a concentration of 1 mg/ml in 10 mM phosphate buffer, pH 7.4, at seven different temperatures in the range 4 -25°C.
Partial Reduction and Alkylation of the Receptor-Partial reduction of receptor heterotetramers from (␣-␤Ј-Z) 2 to ␣-␤Ј-Z was performed essentially according to the protocol of Sweet et al. (20,21). A 5-ml volume of 1 M Tris, pH 10.5, was added to 30 g of receptor in 50 l of 50 mM Hepes buffer, pH 7.6, giving a final pH of 8.5. The mixture was incubated at room temperature for 25 min before the addition of reduced dithiothreitol to a final concentration of 2 mM. Following a 15-min incubation, free thiols of the partially reduced receptor were alkylated by adding iodoacetic acid to a final concentration of 100 mM. The alkylation reaction was allowed to proceed for 60 min at room temperature in the dark. The sample was then desalted on a PD-10 Sephadex G50 column (Pharmacia Biotech, Inc.), preequilibrated with 50 mM ammonium acetate, pH 6.8, and lyophilized. The partial reduction was analyzed using SDS-polyacrylamide gel electrophoresis on a 4 -20% precast NOVEX gradient gel stained using Coomassie Brilliant Blue.
Ligand Binding Assay-Analytical ligand binding assays were performed using phosphate, pH 7.5, 150 mM NaCl, 0.05% Tween 20 as assay buffer and 125 I-labeled IGF-I (Amersham, UK). Pre-IgG-coated Scintistrip microwell plates (Wallac, Finland) were incubated with the 2 S. Strömberg, unpublished observations. FIG. 1. A, schematic representation of the full-length IGF-I receptor; B, the produced extracellular receptor domains fused to the Z affinity handle. The ␣-chain contains two homologous domains, L1 and L2, separated by a cysteine-rich region. The ␤-chain contains extracellular fibronectin type III type repeats (Fn3), a single transmembrane region, and the intracellular tyrosine kinase domain. Black horizontal bars indicate potential ␣-␣ cysteine disulfide bonds. The topology and cysteine arrangement is based on the work by Ward et al. (8). The transmembrane and intracellular domains of the produced soluble receptor are replaced by a single IgG binding affinity handle, Z. The position of the inserted thrombin cleavage site between the receptor and Z is indicated. receptor preparation. Plates were washed three times with PBS-T and three times with assay buffer. One series of dilutions was incubated with 125 I-IGF-I (20,000 -30,000 cpm/0.1 ml). A second series was incubated with 125 I-IGF-I and unlabeled recombinant IGF-I (1 M) for control of unspecific binding. After incubation at 4°C for 3 h or at room temperature for 1 h, plates were rapidly washed with assay buffer and allowed to dry. For competitive ligand binding assays, plates were coated with a defined receptor concentration and washed. Receptors were incubated with 100 l of 125 I-IGF-I in all wells. Second, serial dilutions were made starting with 50 l of 125 I-IGF-I solution containing 3 M unlabeled IGF-I. After incubation, wells were treated as described above. Bound radioactivity was measured using a ␥ counter (Wallac).
Circular Dichroism-Far-UV circular dichroism spectra were recorded using a Jasco J 720 spectropolarimeter. Spectra were recorded from 250 -184 nm at a step resolution of 0.1 nm using a scanning speed of 10 nm/min. Each shown spectrum is the average of three accumulated scans. Protein was dissolved in 15 mM potassium phosphate buffer, pH 7.0, to a concentration of 0.2 mg/ml. Cuvette path length was 0.1 cm. Actual protein concentration was determined using quantitative amino acid composition analysis. Secondary structure contents were estimated using the VARSLC1 variable selection software (22). The CD spectrum of the receptor was compared with a CD library containing reference spectra of 33 proteins. Two proteins were excluded in each iteration, creating 528 possible combinations of the first fit. The two least comparable proteins were removed from the reference set, and the iterative fitting was repeated until the set structure content criteria, more than 96% total secondary structure classified, and the root meant square difference to the reference set Ͻ0.3, were satisfied. Near-UV circular dichroism spectra were recorded from 320 to 240 nm using 0.5-cm cuvettes and a scan rate of 20 nm/min. The protein concentration was 0.3 mg/ml. Spectra shown are the average of five accumulated scans.
Biosensor Analysis-The BIAcore TM , Sensorchip CM5, certified grade, Surfactant P20, and Amine coupling reagents, NЈ-ethyl-NЈ-(dimethylaminopropyl)-carbodiimide, N-hydroxysuccinimide, and ethanolamine hydrochloride were obtained from Pharmacia Biosensor (Sweden). All other buffer chemicals were obtained from Sigma or Fluka (Switzerland). Measurements were performed with IGF-I R -Z as the immobilized acceptor molecule. The receptor was immobilized via primary amine groups as described previously (23) utilizing NЈ-ethyl-NЈ-(dimethylaminopropyl)-carbodiimide/N-hydroxysuccinimide coupling reagents. Protein absorption and coupling was performed at pH 4. Immobilizaton was performed at 5 l/min using 1 ϫ HBS (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.05% P20) as the driving buffer. The final level of IGF-I R -Z immobilization was between 6500 and 7500 resonance units. All kinetic experiments were performed using 1 ϫ HBS as the driving buffer at a flow rate of 8 l/min. The injection of analyte was controlled using the "kinject" command in the BIAlogue TM control software. Ligand sample was injected twice at six different concentrations in random order over the same surface in each series of measurements. Immobilized IGF-I R -Z was regenerated after each cycle by in-jection of 12 l of acidic regeneration solution containing 0.3 M sodium citrate, pH 5, and 0.4 M NaCl. Kinetic measurements were performed by injection of each analyte for 300 s followed by disassociation in buffer flow for 400 s. The temperature in all kinetic experiments was 25°C.
BIAcore Evaluation-The kinetic parameters were calculated using the kinetics evaluation software package, BIAevaluation 2, (Pharmacia Biosensor). The theory of BIAcore measurement techniques and calculations has been extensively described (for a review, see Ref. 24).
Titration Microcalorimetry-The titration microcalorimetry experiments were performed at 15, 25, and 37°C using a 2-ml titration microcalorimetric stainless steel vessel for the multiple channel microcalorimetric system TAM (Thermometric AB, Sweden) (25,26). Another 2-ml vessel, lacking stirring facilities and containing 0.8 ml of water, was used as a calorimetric reference in the twin microcalorimetric unit. The noise level was estimated to be Յ10 nanowatts. Electrical calibration was performed in connection with each experiment, with regard to both the energy and the time constants of the instrument. The calorimetric vessel was loaded with 900 l of 3 M IGF-I R -Z soluble receptor. At each titration, 25-30 aliquots of 4 l of 65 M IGF-I were added with a 6 -7-min interval between each injection (27) using a Hamilton syringe attached with a hypodermic needle mounted on an automated motor-driven pump. To correct for dilution enthalpies, additional dilution experiments where IGF-I was added in the buffer solution alone were performed. The IGF-I R -Z solutions were prepared by exhaustive dialysis against PBS (20 mM sodium phosphate, 145 mM NaCl, pH 7.40) and HBS (20 mM HEPES, 145 mM NaCl, pH 7.40), respectively. The IGF-I solutions were prepared by dissolving freeze-dried desalted stock solution in the same buffer solution that the IGF-I R -Z solutions had been prepared in. The PBS buffer system was used in the experiments at 15, 25, and 37°C, while the HBS buffer system was used in additional experiments at 25°C. The concentrations of the proteins were obtained from amino acid analysis.
The equilibrium constant was estimated by nonlinear regression according to Marquardt-Levenberg (28), where the enthalpy was kept constant at a value arrived at from an individual enthalpy determination of the binding process. In this way it is possible to minimize the number of unknown parameters and neglect any fitting correlation of the fitted parameters.
Thermodynamic Analysis-The binding of a ligand to a protein and protein folding are thermodynamically analogous processes. Except for hydrogen bonding, electrostatic interactions, or other specific interactions, the most important contribution to the thermodynamic properties is the change in hydrophobic hydration. Changes in hydration can be a result of changes in hydration of the binding surfaces of both protein and ligand as well as changes in hydration due to conformational changes of the protein. The heat capacity change, ⌬C p 0 (⌬C p 0 ϭ d⌬H 0 /dT), is the thermodynamic property that is most effected by changes in hydrophobic hydration. When transferring a nonpolar surface from a nonpolar environment to an aqueous environment, e.g. dissociation of protein-ligand complexes or unfolding of proteins, the ⌬C p 0 is large and positive. A striking feature of transferring a small hydrocarbon compound from a nonpolar environment to an aqueous solution is the high FIG. 2. A, outline of the expression vector pKGE978. The receptor nucleotide sequence starts at position 13 and proceeds to position 2842 (amino acids Ϫ30 to 902) (7). Transcription is initiated from the cytomegalovirus promoter region (P/E CMV), shown as a black arrow, followed by the human IGF-I receptor cDNA sequence and the SV40 3Ј-untranslated sequence. Restriction enzyme cleavage sites are shown above the vector outline. B, the translated protein product is shown with signal sequence, amino acids Ϫ30 to Ϫ1, as a white box; the ␣-domain, amino acids 1-740, is shown as a light gray box; and the ␤ extracellular domain, ␤Ј, amino acids 741-932, is shown as a black box. The IgG-binding affinity tail Z is shown as a hatched box. RKRR indicates the intradomain cleavage sequence. The inserted thrombin cleavage site between ␤ and Z domains is indicated. degree of group additivity to the heat capacity (29 -31). The group additivity rules can be translated into models where the thermodynamic properties are functions of accessible solvent surface area, ⌬ASA. Sturtevant (32), Livingstone et al. (33), and Freire et al. (11) have similar approaches to divide the contributions from hydration/dehydration of nonpolar and polar-accessible surface areas, ⌬ASA np and ⌬ASA pol , respectively, by parameterizing ⌬H 0 and ⌬C p 0 . It has been shown that the major contributions to ⌬C p 0 and ⌬H 0 for protein folding/ unfolding or protein-ligand interactions can be rationalized in terms of ⌬ASA (11,32,33).
From the temperature dependence of the enthalpy, the heat capacity was calculated.
Due to the large heat capacity change, the enthalpy and the entropy are strongly temperature-dependent. Equations 2 and 3 show how the enthalpy, ⌬H, and the entropy, ⌬S, vary with the absolute temperature, T, relative to a reference temperature, T ref .
The proton exchange in the binding process, i.e. proton linkage, was obtained by determining the enthalpy of binding in the two different buffer systems, PBS and HBS, at 25°C. The intrinsic binding enthalpy, ⌬H int , and the number of linked protons, ⌬n, is calculated from Equations 4 and 5.
where ⌬H obs1 and ⌬H obs2 are the observed enthalpies of binding in the two different buffers and where ⌬H ion1 and ⌬H ion2 are the ionization enthalpies for the used buffers. When analyzing the calorimetric data we have used the method developed by Freire et al. (11) and the parameters from Xie and Freire (13) for calculating changes in hydration-accessible nonpolar surface area, ⌬ASA np , and accessible polar surface area, ⌬ASA p , upon binding. When corrections have been made for proton linkage, as described by Equations 4 and 5, the enthalpy of binding can be parameterized as follows, and the heat capacity can in analogy be rationalized as follows, The parameters a(T)-d(T) are fitted parameters calculated by Xie and Freire (13) (i) The contribution from the change in hydration upon binding, ⌬S hyd , at temperature T is calculated from the heat capacity change using a reference temperature, T s ; The reference temperature (T s ) is set to 385 K. This temperature is the convergence temperature of the entropy, where the hydrophobic hydration contribution to entropy is zero (34,35). At T s the entropy of unfolding of proteins converge to the same value. This is also the case for the dissolution of small hydrophobic compounds in aqueous solution (35). The convergence temperature has been interpreted as the temperature at which the hydrophobic contribution to the total entropy has happened.
(ii) There are contributions to the total entropy change due to reductions in rotational/translational degrees of freedom, ⌬S rot/trans . Kauzman (36) and Murphy et al. (37) have shown that ⌬S rot/trans is well approximated by the cratic entropy, Ϫ8 cal K Ϫ1 mol Ϫ1 , for 1:1 binding stoichiometry.
(iii) The entropic contribution due to the change in the number of particles in the system, ⌬S no part , is a statistical entropic effect. ⌬S no part ϭ Rln ͩ no. of particles in final state no. of particles in initial state ͪ (Eq. 9) In the initial state for a 1:1 stoichiometry there are two particles, while in the final state there is only one particle. R is the gas constant (R ϭ 1.987 cal (mol K) Ϫ1 . Thus, ⌬S no part is for 1: (iv) In the protein-protein interactions there can be changes in ionization of amino acid residues. This will contribute to the total entropy change as an ionization entropy contribution, ⌬S ion . The magnitude to this contribution is dependent on which type of residue that is involved in the change in ionization.
(v) In the binding process involving a protein, it is likely that there will be changes in conformational degrees of freedom. This can be due to conformational changes in the peptide backbone and/or conformational changes of individual residues. The changes in conformational freedom will contribute to the total changes in entropy, ⌬S conf .
Adding up the contributions (i-v) we arrive at the following expression for the total change in entropy upon binding.

RESULTS
The extracellular portion of the receptor was produced as a soluble secreted molecule with an affinity handle fused to the C-terminal end. The use of the Z affinity tail has been successfully implemented in the production of a large variety of proteins in various prokaryotic production systems (38) and is in this report used in a eucaryotic host system.
Cloning/Sequencing-The receptor was cloned and sequenced in the production vector pKGE 978, and the sequence was found to be fully in agreement with the published sequence (7).
Cell Growth-Human primary kidney 293 cells were transfected for transient expression of the IGF-I receptor. In addition, 293 clones expressing the receptor were selected for establishment of stable cell lines. The selected 293 cell line yielded high and stable expression combined with a minimum of proteolytic degradation.
Protein Characterization-The small scale cell growth used in the initial experiments, up to 50-ml flasks, gave a production yield ranging from 1 to 5 mg of IgG-purified sIGF-I R -Z/liter of conditioned media. The purity of the receptor was determined using SDS-polyacrylamide gel electrophoresis (Fig. 3A). Detection of the Z affinity handle using peroxidase anti-peroxidase staining (Fig. 3B) reveals a single band in the unreduced sample (lane B2) corresponding in size to full-length soluble receptor. In the reduced sample (lane B1), a band corresponding to the size of the ␤-Z-chain is stained. No additional bands, and therefore no detectable degradation products, containing the Z tail were found. The determined amino acid composition corre- sponds to what is expected from theoretical values (data not shown). Five cycles of N-terminal sequencing revealed two different sequences corresponding to a correctly processed ␣-chain N terminus after signal peptide removal and a postcleavage ␤-chain N terminus, respectively.
Receptor processing and association status were analyzed by partial reduction analysis. This analysis is based on the fact that the ␣-␣ interchain (class I) disulfides are more susceptible to reduction than intrachain (class II) disulfides (20). Limited reduction results in the appearance of ␣-␤Ј-Z in addition to the parent (␣-␤Ј-Z) 2 receptor. The gel in Fig. 3A shows the presence of one species in the unreduced sample (lane A2), corresponding approximately to the size of (␣-␤Ј-Z) 2 receptor. The totally reduced sample (lane A1) exhibits two bands of approximately 134 and 70 kDa, corresponding to ␣and ␤Ј-Z chains, respectively. The partially reduced sample (lane A3) shows the presence of two bands corresponding to nonreduced receptor and a band of intermediate molecular mass, 180 kDa, which may correspond to the ␣-␤Ј-Z half receptor. The molecular masses determined may deviate from theoretical values due to changes in gel mobility originating from glycosylation and different levels of intrachain reduction in the samples as well as possible charge differences originating from the alkylating agent.
Glycosylation analysis of the receptor reveals a sugar content of 26.1% by mass. The composition of associated sugars is as follows: fucose, 2.2.%; mannose, 7.7%; galactose, 4.6%; and glucosamine, 9.6%. The sialic acid component was found to be 2%. Total molecular mass based on the theoretical molecular mass, derived from amino acid composition plus sugar content, is 280 kDa for the (␣-␤Ј-Z) 2 receptor. Attempts to determine the molecular mass of the complete receptor assembly using mass spectrometry did not succeed using either laser desorption or electrospray methodology.
Laser light-scattering analysis of both free and ligand-complexed receptor indicates an apparent hydrodynamic radius slightly larger than expected from theoretical calculations. The calculated molecular mass of free receptor at 25°C was approximately 370 kDa, and the mass of the complex was 410 kDa. The determined value may differ up to 20% from the actual value in this analysis. From the light scattering data the receptor was found to be monodisperse, indicating a homogenous receptor population. The difference in size between free and complexed receptor is much larger than the actual size of the ligand (7.5 kDa) and may reflect different deviations from the theoretical globular shape assumed in the calculation of molecular mass.
Circular Dichroism-CD spectroscopy was used to detect possible changes in secondary structure content upon ligand binding. The far-UV spectra of free and ligated receptor are shown in Fig. 4A. The spectrum of the free IGF-I ligand is subtracted from the complexed receptor spectrum. Calculations of the secondary structure components using VARSLC1 variable selection suggest a structure content for the free receptor of 13% ␣-helical, 26% anti-parallel ␤-sheet, 9% parallel ␤-sheet, 18% turn, and 35% unstructured. Spectra of IGF-I R -Z complexed with an excess of ligand (IGF-I) (spectra of the ligand subtracted) reveal small but reproducible differences over the wavelength range suggesting that these differences observed are a direct consequence of binding IGF-I to its receptor. There is a decrease in negative ellipticity for the complex over the entire range. The calculated structural element content of complexed receptor is 14% ␣-helical, 28% antiparallel ␤-sheet, 7% parallel ␤-sheet, 20% turn, and 31% unstructured. The absolute values of secondary structure may contain added uncertainties due to contribution to negative ellipticity from associated oligosaccharides below 210 nm.
The near-UV spectra of the free receptor, of the ligandcomplexed receptor with ligand signal subtracted, and of the sum signal of free receptor and ligand, are shown in Fig. 4B. In contrast to the far-UV spectra, there is an increase in negative ellipticity upon ligand binding in the near-UV spectra. The negative ellipticity at 217 nm is decreased by 9%, and at 286 nm it is increased by 57%. The near-UV spectra is sensitive to aromatic side chain positions and cysteine bridge conformations.
Ligand Binding Assay-The binding displacement of radiolabeled IGF-I to purified IGF-I R -Z, using IGF-I as a tracer, is shown in Fig. 5. The half-maximal inhibiting concentration (IC 50 ) value calculated from the binding curve is 0.5 nM.
Biosensor Analysis-BIAcore kinetic measurements were performed for a series of known ligands to the IGF-I R : IGF-I, IGF-II, des(1-3)-IGF-I, and insulin. IGF-I R -Z was immobilized, and sensorgrams for the different ligand interactions are shown in Fig. 6. A summary of the calculated rate and equilibrium constants is found in Table I. IGF-I affinity was determined in three separate runs using duplicate injections of each concentration, and affinity for the other ligands was determined in one run, also using duplicate injections. The cumulative error in the determined association constants from all runs was estimated to be less than 12%, calculated as the square root of the sum of the squares of the errors in amino acid analysis, pipetting, and data fitting. Determined kinetic parameters (Table I) indicate that differences in both association rates and disassociation rates underlie the observed differences in calculated association equilibrium constants for the tested ligands. The des(1-3)-IGF-I variant has an association rate that is about 1.5 times faster than that of native IGF-I, while the association rate for IGF-II is about 2.8-fold slower than IGF-I. IGF-I and des(1-3)-IGF-I have similar disassociation rates, while that of IGF-II is 2 times faster than that of IGF-I. The insulin association was too slow to be determined, using BIAcore technology at the conditions used. Ligand association was analyzed for insulin at concentrations up to 500 times higher (0.17 mM) than those used for IGF-I. The differences in disassociation rate between the ligands are less pronounced, insulin excluded, than the association rates. The data for insulin disassociation is best fitted to a two-site model in contrast to the other ligands. Out of the two distinct disassociation rates, one is 100 times faster than that of IGF-I, 180 (s Ϫ1 ⅐10 Ϫ3 ) and the other is 10 times faster, 18 (s Ϫ1 ⅐10 Ϫ3 ). The calculated association equilibrium constants for IGF-I R -Z are thus about 2-fold higher for des(1-3)-IGF-I than for native IGF-I, while the value for IGF-II is about 5-fold less, accounted for both by slower association and faster release from the receptor.
Titration Microcalorimetry for IGF-I Binding to IGF-I R -Z-The measured heats of binding were at all temperatures exothermic. The largest heat was obtained at the first injection at each titration series and ranged from 30 to 65 J (Fig. 7). At 15 and 25°C, the measured heat values in the PBS buffer were below the limit of accurate determination of equilibrium constants. However, due to the larger enthalpy at 37°C in PBS and at 25°C in HBS, data could be fitted to a 1:1 binding model with acceptable statistical output (Fig. 8). Other thermodynamic binding models, e.g. models including a complex with stoichiometry of 1:2 or 2:1 (IGF-1:IGF-I R -Z), were also tested to establish the most probable thermodynamic model for the binding process. These other models could be disregarded based on statistical analysis. At 15 and 25°C in PBS buffer, the enthalpies of binding were calculated from total integral heats. The results from the titration calorimetric experiments are summarized in Table II. The results from the analysis of the thermodynamic data are listed in Table III. The thermodynamic properties are dominated by the large negative heat capacity. This is typical for protein-protein interactions in which the hydration of nonpolar residues is reduced. Due to the large heat capacity change, the enthalpy and the entropy are strongly temperature-dependent. As outlined under "Experimental Pro-cedures," the large negative heat capacity change can be explained by dehydration of large hydrophobic surface areas. The large and negative enthalpy also implies a strong temperature dependence of the affinity constant, which indeed can be observed by the difference in affinities at 25 and 37°C. The number of linked protons in the reaction was determined to be Ϫ3.1 at pH 7.4. Important contributions to the entropy change upon binding comes from the solvent, ⌬S hyd , stemming from the burial of hydrophobic groups in the binding process and due to a decrease in the conformational degree of freedom in the IGF-I and IGF-I R -Z molecules. This contribution was calculated to be Ϫ131 cal K Ϫ1 mol Ϫ1 in total, at 25°C. For the IGF-I⅐sIGF-I R -Z system, there is no detailed structural information at hand. Thus, it is not possible to account for any specific ion interactions due to deprotonation of amino acids. Additional structural information can to some extent revise the sizes of the contributions to the different thermodynamic properties. However, since the major contributions to enthalpy and heat capacity stems from changes in hydration of the two proteins, the general features of the contributions to the thermodynamic properties can be judged to be valid within the experimental uncertainties.
According to the analysis of the thermodynamic properties, there is a significant reduction in the degrees of freedom for the proteins upon binding, which is manifested by the large and negative ⌬S conf (Table III). Apart from the binding surface areas of IGF-I and IGF-I R -Z, the large change in total accessible surface area can have its origin in changes in hydration of the proteins due to conformational changes. The accessible surface area for free IGF-I is 4070 Å 2 , based on the NMR minimum average structure (39) and the Lee and Richards algorithm (40) implemented in the program INSIGHT (Biosym).

DISCUSSION
In this paper, we have described the ligand binding properties of recombinant sIGF-I R . The detailed characterization of the IGF-I R ectodomain in terms of ligand binding kinetic properties, secondary structure analysis, and thermodynamic properties, with the finding of specific proton release and receptor conformational changes, as well as the 1:1 receptor:ligand stoichiometry, are important contributions to the understanding of the IGF-I molecule signaling mechanism and specificity.
The titration microcalorimetry-derived thermodynamic parameters, and in particular the large entropy loss upon ligand binding, are consistent with a conformational stabilization in-duced by ligand binding. Part of this stabilization could be induced by a conformational change. The titration microcalorimetry data presented here suggest that some 4700 Å 2 get depleted of water when IGF-I binds to IGF-I R Z. This can be either through the burial of binding surfaces on both the ligand and the receptor or from structural changes leading to a reduction of solvent exposed surface in either of the molecules. Assuming rigid molecules, this would mean that more than half ((4675/2)/4070 ϭ 0.57) of IGF-I gets buried upon binding. This is an unlikely large portion of the molecule, supporting the hypothesis that a conformational change does occur in the receptor. The CD analysis of the IGF-I receptor ectodomain suggests that the secondary structure of the receptor changes upon ligand binding. The far-UV CD spectra of free and IGF-I-ligated IGF-I R -Z show small changes in negative ellipticity (Fig. 4), and the near-UV spectral changes are somewhat larger, but in both ranges the changes are reproducible and significant. The difference in the spectra between the complex and the sum of the signals of free receptor and free ligand indicates that changes in position of aromatic side chains and, possibly, cysteine bridge dihedral angels occur in the receptor ectodomains upon ligand association. These changes could be due to secondary structural changes, quaternary structural changes, or both (41).
The three-dimensional structure of the insulin receptor cytoplasmic tyrosine kinase domain has been determined using protein crystallography (6). These data suggest that the activation of the tyrosine kinase involves intradomain transphosphorylation made possible by induced close contact of the domains and structural rearrangements in the protein kinase active site region. Thus, it seems possible that signal transduction across the membrane is the result of a ligand-induced quaternary structural change. The nature of this change may, for instance, involve a scissors-like movement of the receptor chains, a twist of the chains, or a combination of these movements. Binding of ligand to both IGF-I and insulin receptors is proposed to involve negative cooperativity (42). The mechanism has been suggested to involve the presence of both high and low affinity binding sites on each ␣ subunit. A ligand might bind to a high affinity site on one subunit inducing conformational changes enabling the ligand to bridge the two subunits by also binding to the low affinity site on the second subunit. Our studies on IGF-I R -Z give a stoichiometry of 1:1 (IGF-I:(␣-␤Ј-Z) 2 ) with high affinity, which could include two separate binding sites.   The relative potency of binding to the produced IGF-I receptor for the different ligands is in accordance with previous investigations performed using competitive ligand methods (43). The des(1-3)-IGF-I exhibits a faster association rate than native IGF-I and thus a higher affinity. IGF-II has 5 times lower affinity than IGF-I due to both slower on rates and higher off rates. The observed discrepancies in affinity constant between the BIAcore methodology and the titration microcalorimetry are in accordance with previous studies and may be due to higher degree of conformational freedom for the IGF-I R -Z molecule in solution than when it is immobilized on a surface matrix. This reduction in conformational degrees of freedom upon surface immobilization would increase the apparent affinity for the ligand. The receptor binding properties of a number of IGF-I mutants have been tested using both BIAcore technology and classical radiolabeled competition assays. 3 Such studies show that although the absolute value of binding may vary between used methodologies, a series of molecules will have the same relative order of affinities, disregarding the method used. Similar conclusions can be made also for the interaction of human growth hormone mutants to the human growth hormone receptor (44) and for binding of Z mutants and their binding to IgG (45) using displacement and BIAcore methods. In addition, the validity of association equilibrium constants determined by BIAcore kinetic analysis has been addressed in works by Ladbury (51) and Morton et al. (52); in both works the combination of biosensor methodology and isothermal titration calorimetry was used to characterize receptor-ligand interactions. Ladbury and co-workers (51) found that only steady state measurements were reliable in the biosensor setup used. In the work by Morton et al. (52), good agreement was found between kinetic and steady state biosensor measurements as well as titration calorimetry-derived val-ues. We believe the presented kinetic data to be correct based on the following. The association is between a small monomeric ligand and a large immobilized monodisperse receptor, thus minimizing the possibilities of avidity effects, using self-associating proteins or mass transport limitations, that occur when a large receptor associates with a small immobilized ligand. The plot of k s against concentration, used to determine the association rate (k on ), was linear in the used concentration range. The disassociation rate does not change with concentration in the highest concentrations used, and the addition of free IGF-I-binding IGFBP-1 protein in the disassociation phase does not increase k off , demonstrating that detectable rebinding does not occur. The obtained rate information provides the opportunity to more fully quantitate the behavior of the ligands and the effects of introduced mutations in IGF-I R ligands. The association rate of insulin binding to IGF-I R -Z is too low to be determined using the BIAcore; it is possible, however, to calculate the disassociation rate constant from the data. Interestingly, in contrast to the other ligands, a two-site model gave the best fit for insulin disassociation. The rates of disassociation are approximately 10 and 100 times faster, respectively, for the two insulin off-rates than the corresponding rate for IGF-I. A hypothetical explanation for this difference is that IGF-I binds to the IGF-I receptor and induces a conformational change by simultaneously bridging the high and low affinity sites (42). Insulin, on the other hand, binds to the receptor, one molecule to each ␣ subunit, with significantly lower affinity but is unable to induce conformational changes and is therefore rapidly disassociated from these two sites. Analysis of ligand binding in the presence of different antibodies directed toward the IGF-I receptor (46) shows that signal transduction may be blocked independently of ligand binding. This might be the case if the antibody would bind outside the ligand binding site and block a conformational change, thus inhibiting further signal transduction. In the study by Soos et al. (46) there is a further observation that antibodies potentiating IGF-I ligand binding also increase the affinity of insulin for the IGF-I receptor up to 50-fold. These findings are consistent with the thought that insulin on its own is unable to bridge between the ␣-subunit ligand-binding regions of the IGF-I receptor. The antibody could possibly switch the receptor conformation to an "insulincompetent" state by locking the subunits in a conformation allowing formation of an insulin high affinity receptor complex.
It has previously been shown by Gual et al. (47), using conformational sensitive antibodies, that the full-length IGF-I receptor undergoes an autophosphorylation-induced conformational change in its cytoplasmic ␤ region, which is distinctly separate from ligand-induced conformational changes. In this work we have presented data supporting a ligand-dependent conformational change in the extracellular domains of a kinase-deleted IGF-I receptor. Two electron microscopy studies of the structure of the highly homologous insulin receptor have previously been published (48,49). Christiansen and co-workers (48) studied the structure of full-length receptor from placental membrane preparations, both in the presence and absence of insulin. They found two different forms of the receptor, one shaped like an elongated T and a second more X-like in structure. The addition of insulin did not yield a single confor-  mation, but the possibility could not be excluded that structural changes may accompany ligand binding and receptor phosphorylation. Schaefer et al. (49) studied the insulin receptor extracellular domains, using a construct analogous to the extracellular IGF-I R presented in this work and found forms of structures similar to those found by Christiansen and collaborators. These studies support the theory of large flexibility of the receptor ␣ domains. The speculation that the conformations found in these reports represent the structures of receptor with or without ligand present is therefore appealing. Further, Schaefer et al. (49) applied CD analysis to the insulin receptor and concluded that changes in secondary structure content occurred when insulin was added. Detailed characterization of ligand binding of extracellular receptors requires a convenient way to produce sufficient amounts of the mature protein. We have demonstrated the successful use of an affinity handle fusion protein expression system for the production of large amounts of a soluble form of IGF-I R . Purification using the Z:IgG affinity system has high specificity and gives an essentially pure receptor preparation using a single-step chromatography procedure (Fig. 3). The IgG(Fc)-binding Z protein affinity handle, derived from the B domain of Staphylococcal protein A, has been extensively used in a large number of biotechnological applications for the production and purification of recombinant proteins (38). Growth impairments or toxicity effects have not been detected as a result of the production of this prokaryotic protein sequence in the 293 cell line used. The produced receptor variant displays limited, if any, proteolytic degradation even when harvested after prolonged cell growth.
The data presented in this paper support the hypothesis that extracellular IGF-I receptor domains undergo a conformational change upon ligand binding that includes changes in both the amount of accessible surface area and the amount of secondary structure content. The release of protons in the binding process may originate from formation of salt bridges between receptor and ligand. It has been shown by mutational analysis that the charged IGF-I C domain residues are important for high affinity IGF-I receptor binding (50). The proton release could also be part of the conformational change in the receptor structure. These two different possibilities cannot be distinguished without detailed structural or mutational analysis.
In this study, we have shown that the produced soluble IGF-I R -Z binds one IGF-I molecule with high affinity and is capable of structural changes, possibly those involved in the activation of signal transduction. The lack of transmembrane and cytoplasmic regions does not destroy high affinity ligand binding capacity.